Mining Association Rules from Multidimensional Transformer Defect Records

There are various types of transformer device defects and the formation reasons are complex. Exploring the influencing factors and occurrence of transformer devices defects is a focus in the field of power transmission and transformation devices state inspection and evaluation. This paper proposes an analysis method, multidimensional FP-Growth algorithm (MDFPG) to mine association rules from multidimensional transformer defect records. The method combines records from different system of power grid to construct multidimensional records first. Then, the records are preprocessed and encoded into single dimension form. The MDFPG method speeds up the mining process by adding a pruning step. Experiments show that MDFPG method has a better performance than FP-Growth algorithm on large data sets. Some conclusions can be learned from the experiment result, which has a certain value for making equipment maintenance plans and exploring defect occurrence regularity.